Magnetic resonance imaging (MRI) is a rapidly advancing technology which provides new opportunities for the enhanced visualization of anatomic, physiologic, and pathologic features by using a single diagnostic imaging modality. This technology is based on the principle that protons within body tissues and fluids can absorb and then subsequently emit radio frequency (RF) signals when placed in a strong magnetic field. It is the detection of the emitted RF signals from three-dimensional tissue voxels that provides data for the spatial location and contrast discrimination of specific tissues, which are ultimately displayed as pixels comprising a two-dimensional MR image. The intensity of each individual pixel is determined by several biophysical characteristics of the tissue within the voxel. These characteristics include longitudinal relaxation rate (T.sub.1), transverse relaxation rate (T.sub.2), proton density, and flow velocity and direction. To emphasize contrast patterns of specific tissues, different image acquisition parameters or pulse sequences are utilized to produce various types of "weighted" images. The selection of different pulse sequences allows for the generation of spin echo images that are T.sub.1 -weighted, T.sub.2 -weighted, or proton density weighted. Furthermore, gradient echo pulse sequences, or "fast scans", can be utilized to obtain images that possess additional unique tissue-contrast patterns. In each type of image, individual tissues appear differently based on their own inherent biophysical characteristics.
Because the intensity characteristics of individual tissues in different types of MR images are dependent on the pulse sequence parameters selected for image acquisition, various persons have attempted to develop both gray-tone and color display methods for tissue characterization, based on pattern recognition or multispectral analysis techniques using multispectral MR image sets. In color cathode ray tube (CRT) systems and in color raster graphics, the process of additive color mixing creates a spectrum of colors through the superimposition of red, green, and blue visible light regions. Additive color mixing has also been utilized in the National Aeronautic and Space Agency (NASA) LANDSAT satellite imagery system to generate color composites from multispectral image sets. In this system, multiband infrared images are remapped to each of the three primary colors and then superimposed on one another to form a single image possessing a spectral scheme that is based on additive RGB color combinations of spatially aligned pixels. Because of their multiband nature, this same type of image processing has been utilized to generate RGB composites from sets of different types of MRI. Advantages include the potential for a more realistic appearance in computer-generated images, enhanced ability of the viewer to interpret different types of data present in an image, and since color images may display variable hue, saturation, and brightness values much more information may be conveyed to the viewer.
Vannier et al have demonstrated computer-generated color displays by producing classified images with enhanced discrimination of different body tissues and fluids, as set forth in Multispectral Magnetic Resonance Image Analysis, (Crit. Rev. Biomed. Eng., 15:117-144, 1987). Vannier has produced color composites by assigning red or green or blue (RGB) to two or three channels to produce the composite. A problem with this type of system is that the arbitrary and simple direct assignment of one image to red, another to green, and a third to blue, limits the channels to three, and usually does not provide desired visual characteristics to allow for semi-natural tissue appearances. Furthermore, Koenig et al in Pattern Recognition for Tissue Characterization in MR Imaging (Health Care Instrum., 1:184-187, 1986) investigates the area of pattern recognition for tissue classification. Koenig et al classified four types of tissue by pixels defined by a feature vector which contained information from computed MR parameters. These parameters were represented by gray value, neighborhood relations and texture. The image acquisition parameters or pulse sequences used were non-standard or not the routine protocol for brain MR imaging. In Koenig's methods, multiplication of a feature vector by different functions maps a pixel into a decision space so that pixelwise classification is accomplished by the decision for the maximum components of the estimation factor. In other words, this is a statistical classification scheme which creates high contrast, gray-tone masks which map the highest probability class based on a priori training of a classifier function. Such classification methods are important and hold promise but are generally not clinically feasible due to the requirement for training the classifiers, the inherent problems of "computer diagnosis" and consequent physician non-acceptance. Since it segments tissue types by a strictly mathematically method, there is a potential for misclassification of volume averaged or continuous gradient intensities which are common in MRI. Kamman et al in Unified Multiple-feature Color Display for MR Images (Magn. Reson. Med., 9:240-253, 1989) has proposed a color display method based on calculated T.sub.1 and T.sub.2 relaxation times, as well as the proton density, of particular tissues of interest. Color images that simultaneously represent both T.sub.1 and T.sub.2 relaxation times were generated by mixing the primary colors of R, G, B. However, a color scale is required to interpret color "code" and this method is limited to three channels, T.sub.1 and T.sub.2 comprised of images which are not in the same visual format (calculated images) as would be viewed for standard and routine MRI diagnostic evaluation.
In U.S. Pat. No. 4,690,150, issued Sep. 1, 1987 in the name of Mayo, Jr., a method of producing a color overlayed image taken from a monochrome image is described. The method includes the steps of scanning for the image, i.e., MRI, storing the image in memory, and thereafter filtering the image by utilizing a dividing circuit and an average value circuit. The divided image along with the background information is sent to several ROMs associated each with the red, green and blue outputs. The ROMs include look-up tables which determine the values of the red, green and blue. Thereafter, the output signals from the RGB are sent to digital to analog converter for display on a CRT monitor. The patent produces a pseudocolor component thematic mapping limited to two channels in which a second image treatment is modulated by its pixel values' relationship to corresponding pixel values of the first image. The pseudocolor images produced provide two images which convey to different parameters.
A second U.S. Pat. No. 4,789,831 issued Dec. 6, 1988 in the name of Mayo, Jr. describes a similar method to his previously mentioned method in that a first image is used as an intensity image while a second image is used to tint the first image. This hue being determined by the sign and magnitude of deviation of the second from the first image. This method appears to be limited to two images (T.sub.1 and T.sub.2), does not produce seminatural color assignments and may create visually confusing color images.
U.S. Pat. No. 4,998,165, issued Mar. 5, 1991 in the name of Linstrom discloses a method of selectively changing monochrome color signals to color in diagnostic imaging. The method includes receiving the video signal of a monochrome image, stripping- the synchronizing signal therefrom, and sending same through an analog and digital converter for digitization of each pixel. The signal is thereafter sent to three memories each associated with red, green and blue and each comprising a different look-up table depending upon the magnitude of the monochrome image. The look-up tables assign a value to each of the pixels depending upon whether it is a low, medium or high magnitude in the video signal, i.e., intensity. The output of the RGBs are sent to a digital to analog converter thereafter to a monitor. The Linstrom patent is basically a pseudocolor method to apply color to a single achromatic or gray-tone image so that the single image pixel intensity values have been assigned colors. The application of this method appears to be primarily for single parameter modalities. It is therefore limited to a single channel or image.
Although the color display methods previously described have yielded images with enhanced differentiation of various tissues and fluids, most of the images that have been generated possess unnatural color combinations that often require special legends or understanding the often complex algorithms used in order to facilitate their interpretation by the viewer. Furthermore, many of the methods produce either pseudocolor or arbitrarily assigned false color composites. The pseudocolor is applied to single parameter images, and produces unnatural boundaries or contours non-existent in the original image based upon applying color by using ranges of pixel values. In false color, when R, G, B are each assigned to an image the following results are expected.
If a spatially corresponding pixel has a high intensity on all images, the composite pixel is white, if one pixel is high and red, a second is medium intensity and green, and a third is dim and blue the false color composite will be orange, etc. The composite color is simply based on the linear combination of the component monochrome pixels' hues, saturations and intensities (colors) into a resultant voxel.
The problem with pseudocolor images is that with multiple pulse sequence MR the difference in pixel value ranges within one image conveys only spatial information for that "weighting". Using the intensity of a pixel as an address for color look-up table assignments produces a single parameter image displayed as sharply designated regions indicating some finite number of divisions of pixel intensity range. This is fine for single parameter displays such as thermography. However, this creates visually confusing spatial images with artificial contours. Furthermore, reference to a color code table would be required for interpretation as in thermography. For there reasons, pseudocolor methods do not produce semi-natural, visually coherent, or intuitively transparent color images. The artificial graded contours in what should have been smooth gray-tone transition introduces false and misleading spatial information. This is very troublesome in diagnostic radiology.
Additionally, the information which is intended to be conveyed is the relative intensity of corresponding pixels in each of the channels or images of the multiparameter set. This point is well appreciated in the prior art as multispectral analysis, pattern recognition and other statistical classifications algorithms. These are important and appropriate applications for multiparameter data. However, the same introduction of artificial contours, the subjectivity to inhomogeneous field strength and other instrumental variations pose very significant problems for a strictly quantitative statistical classification approach to providing a thematic map representing specific tissues and fluids.
Advantages of color image display, in contrast to achromatic or monochromatic presentation, include the potential for a more realistic appearance, enhanced information processing, and increased ability of the viewer to discriminate and interpret related and unrelated data. Moreover, while each display point in a gray-tone image possesses intensity as its only variable, each display point in a color image has three variable attributes hue, saturation, and intensity, which allows greater information capacity within the display.